Real-Time Macro Information and Bond Return Predictability: Does Deep Learning Help?
Preprint
- 10 January 2020
- preprint
- Published by Elsevier in SSRN Electronic Journal
Abstract
This paper examines whether deep/machine learning can help find any statistical and economic evidence of out-of-sample bond return predictability when real-time, instead of fully-revised, macro variables are taken as predictors. First, we find some statistical evidence for forecasting non-overlapping excess bond returns using the deep learning models. Second, for forecasting overlapping excess bond returns, we find more apparent statistical evidence, for which the two nonlinear regression tree models, boosted tree and random forest, outperform the deep learning models. However, this statistical evidence is weaker than that from using fully-revised macro data and cannot generate any economic gains for a mean-variance investor, regardless of her level of risk aversion and whether she can take short positions.Keywords
This publication has 33 references indexed in Scilit:
- Term Premia and Inflation Uncertainty: Empirical Evidence from an International Panel DatasetAmerican Economic Review, 2011
- Frontiers of Real-Time Data AnalysisJournal of Economic Literature, 2011
- Macro Factors in Bond Risk PremiaThe Review of Financial Studies, 2009
- On the Volatility and Comovement of U.S. Financial Markets around Macroeconomic News AnnouncementsJournal of Financial and Quantitative Analysis, 2009
- Time-Varying Risk Premiums and the Output GapThe Review of Financial Studies, 2008
- Predicting Excess Stock Returns Out of Sample: Can Anything Beat the Historical Average?The Review of Financial Studies, 2007
- Approximately normal tests for equal predictive accuracy in nested modelsJournal of Econometrics, 2007
- Beliefs, Doubts and Learning: Valuing Macroeconomic RiskAmerican Economic Review, 2007
- Bond Risk PremiaAmerican Economic Review, 2005
- Asymptotic Inference about Predictive AbilityEconometrica, 1996